[maker-devel] SNAP scores and AED scores

[Carson Holt]

On a side note.  Because the MAKER models involve modifying either the ab
initio SNAP model or manipulating the underlying scoring scheme using
hints, the SNAP score on those is virtually meaningless.  However Ian Korf
has developed a tool that can take any gene structure and reverse generate
a score (i.e. what would the score of this gene have been if SNAP would
have called it that way in the first place).

I believe the tool is called fathom and is part of the SNAP package.  It
is not well documented, so you might have to contact Ian Korf directly for
that.  You can use the maker2zff tool to generate the input to fathom.

Thanks,
Carson


On 2/13/14, 1:11 PM, "Carson Holt" <carsonhh at gmail.com> wrote:

>No.  Snap genes do not disappear. All SNAP ab initio calls will always be
>kept as reference fetters marked snap_masked (for repeat masked genome)
>and snap (for unmasked genome).  MAKER then runs SNAP another time where
>it feeds hints to SNAP based on EST and protein alignment evidence.  These
>hint based models can then compete against the ab initio SNAP models to be
>promoted to genes if their AED scores are better.  Fianl models can also
>get UTR added based on EST evidence.  That is why you can get models from
>MAKER that do not match the original SNAP ab initio calls.
>
>So in summary, all SNAP ab initio models will be in snap_masked.  The
>MAKER models will consist of hint based SNAP rerun plus SNAP ab intio
>models processed to add UTR.
>
>Thanks,
>Carson
>
>
>
>On 2/13/14, 3:17 AM, "Masa Roller" <masa at bioinfo.hr> wrote:
>
>>Dear all,
>>
>>I ran snap2 based gene prediction through maker.
>>
>>In the resulting gff file, in the source "snap_masked" I can find the
>>score in the score column of every snap prediction that did not get
>>promoted to a maker gene. This would be the score of how well the
>>prediction matches the HMM?
>>
>>It seems to me that those snap models that are given gene status no
>>longer appear as snap_masked source but only as source "maker". Maker
>>then removes the score column, instead giving AED and eAED scores (which
>>are more about how the model corresponds to the evidence). When viewing
>>the maker transcripts and SNAP predictions in a browser, they do not
>>match (mostly, maker predictions are longer).
>>
>>I am interested in the score of individual gene predictions that
>>underlined maker gene models. Where could I find that information?
>>
>>Many thanks!
>>
>>_______________________________________________
>>maker-devel mailing list
>>maker-devel at box290.bluehost.com
>>http://box290.bluehost.com/mailman/listinfo/maker-devel_yandell-lab.org
>
>